Zhenzhen API Settings (Legacy)

If you loaded a Zhenzhen workflow and this node appeared, nothing is broken. Zhenzhen_api_set - displayed as "Zhenzhen API Settings (Legacy)" - is a subclass of the pack's current T8Zhenzhen_API_Settings node, flagged DEPRECATED = True, and kept registered under its old class name purely so workflows saved before the rename still load. It is, bit for bit, the same node.

What it does

Identical schema, identical behavior. Inputs are api_base (dropdown: seedance_low_price → the budget https://api.seedance.nz site, zhenzhen → the overseas https://ai.t8star.org, ip → a custom endpoint), apikey, optional custom_ip and skip_error. Outputs are the apikey string pass-through and the api_config object (ZHENZHEN_SEEDANCE2_CONFIG) that you wire into any generation node's api_config input. It makes no API call itself - it just bundles your key and chosen base URL into a config dict for the Seedream, Seedance, Sora, Veo, Suno and Midjourney wrappers downstream.

The pack also registers an even older alias, Comfly_api_set, from before the pack was renamed away from the "Comfly" branding. Same story there.

Why it exists

The author renamed the canonical settings node to T8Zhenzhen_API_Settings, and to avoid breaking every saved workflow the legacy class stayed in the registry as a compatibility shim. The pack's frontend extension goes one step further: when you load an old graph it silently migrates legacy settings nodes to the new class name before the workflow runs, which is why you'll often see this node in the JSON but not on your canvas.

So: found it in a downloaded workflow? Let it be, it works exactly like the current node. Building something new? Use Zhenzhen API Settings (T8Zhenzhen_API_Settings) and you'll get the same thing without the "legacy" baggage and with the signup button the extension attaches to the canonical node.

Install and gotchas

Same as the pack: ComfyUI Manager → search Comfyui-zhenzhen, or git clone https://github.com/T8mars/Comfyui-zhenzhen into ComfyUI/custom_nodes, then restart. No models download, but the pack drags in a heavy dependency tree (openai, transformers, opencv-python, torchaudio, …) even though this node only needs to assemble a dict.

Worth knowing regardless of which settings node you land on: the overseas zhenzhen channel serves from overseas servers, so mainland users hit 443 and need a VPN (author's own README warning), and empty or wrong keys produce "Seedance API key is required…" on the generation node, not here. It's a fan-run API reseller - closed models only, per-call cost, your key sits in the workflow JSON, so don't share saved workflows with it still in there. The KB's external-api-nodes.md has the broader argument for when this whole approach makes sense.